1. Home
  2. Knowledge Base
  3. O - Skepticism and Debunking
  4. 098. Methodologies for Debunking
  5. Skeptic’s Toolbox: Photogrammetry & Image Analysis

Skeptic’s Toolbox: Photogrammetry & Image Analysis

Since 2020 the U.S. Department of War and intelligence community have acknowledged that some airborne anomalies are real unknowns and merit structured investigation. Both NASA’s 2023 Independent Study Team and DoD’s All-domain Anomaly Resolution Office (AARO) call for higher-quality data, standardized reporting, and rigorous analysis. That’s exactly what photogrammetry (measuring from images) and forensic image analysis deliver. (U.S. Department of War)

What you’ll learn here: how to preserve media, extract trustworthy metadata, reconstruct camera geometry, estimate ranges/sizes/speeds, geolocate and time-stamp scenes, and stress-test claims against physics, air/space traffic, and known imaging artifacts, without needing a laboratory.

Golden Rule: Preserve the original

  1. Make a verified copy before you look.
    • Compute hashes (e.g., SHA-256) and log them with date/time and examiner.
  2. Never “Save As…” the evidence file. Use copies for working analysis.
  3. Document chain-of-custody.

Standards to cite in your workflow: SWGDE (Scientific Working Group on Digital Evidence) Guidelines for Forensic Image Analysis and Best Practices for Image Authentication, they outline evidence handling, documentation, examiner qualifications, and bias controls. Keep PDFs alongside your case notes. (SWGDE)

Rapid triage

Reverse-search & duplicates

Source hygiene & context

  • Use open-source verification playbooks from Bellingcat and the Verification Handbook to check uploader identity, posting history, and cross-platform trails. Bookmark their checklists. (bellingcat)

Quick EXIF/metadata sweep

  • ExifTool (CLI) gives a fast, complete read of embedded metadata; export JSON or CSV for your report.
  • exiftool -a -u -g1 -json suspect.jpg > exif.json
  • exiftool -FileName -FileSize -MIMEType -CreateDate -ModifyDate -GPS:all -XMP:all suspect.jpg

Note: Metadata can be missing or edited; treat it as claims. Cross-check time, camera, and location against harder signals (see below).

Metadata forensics

JPEG Quantization Tables (QTs)

  • Cameras and editors often leave distinctive JPEG quantization tables. Comparing a file’s QT against known signatures can support (or contradict) EXIF camera claims and detect double-compression. Tools: JPEGsnoop (free) and commercial suites like Amped Authenticate. (GitHub)
  • Research basis: Kornblum (DFRWS 2008), Farid et al. (2011) show how JPEG headers (QTs, Huffman tables, thumbnails) can form a “camera/editor signature.” Use this to corroborate the source, not as sole proof. (DFRWS)

PRNU (sensor pattern noise)

  • In advanced cases, Photo-Response Non-Uniformity (PRNU) can link an image to a specific sensor across a set of photos (works best with many images from the same device and minimal processing). See Lukas–Fridrich–Goljan (2006). Use with caution on social-media-compressed content. (ResearchGate)

ELA caveat

  • Error Level Analysis (ELA) is popular on the web but easy to misinterpret; treat ELA as a triage visualization, not a verdict. See technical critiques noting false positives/negatives. (Semantic Scholar)

Single-image photogrammetry: getting size & distance from one frame

When you have only one picture, you can still bound geometry:

Get the camera model right

  • Extract/derive focal length in pixels:
    fpx = fmmsmm×Wpxf_{px} = \frac{f_{mm}}{s_{mm}} \times W_{px} (approx.; better is full calibration).
  • If the lens is fisheye/ultra-wide, use the proper model to avoid over/under-estimating angles. OpenCV provides pinhole, fisheye, and omnidirectional models, with tutorials to calibrate from a checkerboard or ChArUco board. (OpenCV Documentation)

Angular size method

  • Measure the object in pixels (pp) and field of view in radians from fpxf_{px}. The subtended angle θ≈p/fpx\theta \approx p/f_{px} for small angles.
  • If you have any reference of known size (e.g., building panel, street lane width), compute a scene scale and infer distance or object size.

Shadow/solar geometry

  • If shadows are visible on level ground, the object height HH ≈ L⋅tan⁡(α)L \cdot \tan(\alpha), where LL is shadow length and α\alpha is solar elevation at the scene’s time/location.
  • Get authoritative solar azimuth/elevation from NOAA’s Solar Calculator (log the exact inputs and output). (NOAA Global Monitoring Laboratory)

Perspective & vanishing lines

  • Converging verticals/horizontals let you recover camera tilt/roll; with two or more vanishing points you can back-solve for metric information (classic single-view metrology). OpenCV has primitives for homographies and pose from planar patterns if you can identify a rectangle in the scene. (OpenCV Documentation)

Deliverable: a worksheet (spreadsheet is fine) recording fpxf_{px}, pixel measurements, solar data, assumptions, and outputs with uncertainty ranges.

Multi-image photogrammetry: triangulate, don’t speculate

Whenever you have more than one photo (or a short burst), reconstruct in 3D:

GUI pipeline (no code): Meshroom (AliceVision)

Power users: COLMAP / OpenMVG

  • COLMAP (GUI/CLI) and OpenMVG (with OpenMVS) are gold-standard research pipelines for SfM/MVS. Use them when you need full control (feature types, bundle adjustment settings, rig modeling, etc.). (GitHub)

Minimal two-view triangulation

  • With two vantage points and a known baseline (or recovered via SfM), triangulate the object’s 3D point(s). If the object is moving, track a feature across frames and triangulate per time to get a motion track.

Deliverable: a reconstruction report: software versions, input set, camera models, reprojection error, and the measured object distance/size with error bars.

Video forensics & kinematics: speed, wobble, and the rolling-shutter trap

Instant basics

  • Frame rate & exposure: Extract exact fps and shutter/exposure per frame when available (many phones vary exposure).
  • Motion blur → angular rate: Blur length in pixels bb ≈ angular speed ω\omega × exposure time tet_e × fpxf_{px} / distance. Even rough bounds help adjudicate “instant accelerations.” (PMC)

Rolling shutter artifacts

  • Most CMOS phones/tablets scan rows top-to-bottom. Fast panning or fast scene motion bends/tilts straight lines (“jello”) and can mimic impossible accelerations. Learn the tell-tales and estimate the row-readout time to predict artifact magnitude. See overviews and classic correction work (Baker et al., Microsoft Research). (Photonics.com)

ENF & lighting fingerprints (advanced)

  • AC mains flicker (50/60 Hz) sometimes imprints Electrical Network Frequency traces in video, useful for authenticity and time/place constraints, with caveats for rolling shutter. (IEEE Signal Processing Society)

Deliverable: a kinematics note estimating angular rates/speeds with a candid artifact analysis (did RS artifacts or autofocus pulsing bias the estimate?).

Geolocation & timing: anchor the scene in the world

Land features & map matching

  • Follow Bellingcat’s geolocation workflow: horizon profiles, road furniture, signage, terrain shape, shadows. Log every positive and negative match. (bellingcat)

Sky & star field

Satellite passes

  • Cross-check bright passes with Heavens-Above and authoritative TLE/GP data from CelesTrak (log the epoch!). This can eliminate many “lights that pace the horizon” or glints. (heavens-above.com)

Aircraft & balloons

  • Query historical ADS-B for the time window (ADS-B Exchange) and check near-surface winds with weather reanalyses; combine with balloon NOTAMs when available.

Sun & shadow timing

  • Re-derive local solar azimuth/elevation with NOAA Solar Calculator to validate claimed capture times (use absolute timestamps like “2025-08-07 20:13 PDT”). (NOAA Global Monitoring Laboratory)

Deliverable: a geolocation packet with map screenshots, sky simulations, TLE snapshots, and an evidence log.

The usual suspects: optical & contextual confounders

  • Bokeh misidentification: Out-of-focus point sources near infinity become lens-aperture-shaped blobs that appear to “morph.” Test: refocus to infinity; stop down; compare bokeh shape to aperture blades.
  • Insects/birds near lens: Large parallax, erratic path, “transparent” wings; often cross the FOV in <1 s. Test: high-shutter re-shoot at same spot; check scale vs. known background at different focus distances.
  • Lens flare/internal reflections: Ghost images line up with bright sources at specular angles; path locked to camera move, not sky. Test: block the light source; see if the “UAP” dies with it.
  • Satellites/Starlink trains/ISS: Predictable passes; brightness flares near specular points. Test: Heavens-Above/CelesTrak time-correlation. (celestrak.org)
  • Drones/balloons/lanterns: Speeds, wind drift, LED patterns; ADS-B for some drones; winds aloft for balloons.
  • Rolling-shutter illusions: Sudden “jumps” coincident with fast pan/jerk; wobbly skyline. Test: analyze row time and match skew direction. (Microsoft)

Casework template

A) Intake & preservation

B) Provenance & duplication

  • Reverse-search; account history (Bellingcat checklists). (bellingcat)

C) Metadata

  • ExifTool dump; QT signature check (JPEGsnoop / Amped Authenticate). (ExifTool)

D) Geometry

  • Determine camera model (OpenCV model choice), get fpxf_{px}. Single-image bounds or SfM (Meshroom/COLMAP). (OpenCV Documentation)

E) Kinematics

  • Video fps/exposure; motion-blur rate; rolling-shutter analysis. (PMC)

F) Geolocation & timing

  • Landmarks; sky solve; satellite/airspace/wind cross-checks; NOAA sun. (heavens-above.com)

G) Hypothesis testing

  • Construct the strongest mundane hypothesis and attempt to falsify it with your measurements. If it survives, report it honestly; if it fails, upgrade the claim taxonomy (see §13).

Worked micro-example

Claim: “Spherical, glowing UAP hovering over an oil terminal; cell-phone video at dusk; accelerates ‘impossibly’ to the east.”

Steps in brief

  1. Triage: Reverse-image search: no prior matches. Source is a local worker with corroborating post history.
  2. Metadata: EXIF present but timezone looks off by −1 h (common). JPEG QTs match the stated phone model’s signature range; no double-compression detected. (farid.berkeley.edu)
  3. Geolocation: Storage tanks and lattice tower line up with a known terminal on the coast; horizon and road layout confirm.
  4. Timing: Dusk in video looks like civil twilight; NOAA solar az/alt at “2025-06-12 21:05 local” matches the sun direction seen in reflections; uploader’s timestamp likely auto-adjusted; we log both values. (NOAA Global Monitoring Laboratory)
  5. Kinematics: Frame-level analysis shows ~1/120 s exposure at 30 fps; blur trails on the “sphere” align with camera pan. Rolling-shutter row time estimated ~10 ms; observed skew in light poles fits expected RS distortion. The “impossible acceleration” coincides with a fast rightward pan, artifact. (Microsoft)
  6. Air/space traffic: Heavens-Above shows a bright satellite pass ~21:07 with magnitude ~2.4 low in the east; track overlays the object’s path; specular flare timing matches the brightening. TLE snapshot archived. (heavens-above.com)
  7. Conclusion: Misidentification – satellite flare amplified by rolling-shutter/pan illusions. Report filed with all measurements, links, and reproducible steps.

What this approach does for UAP studies

  • Filters noise without dismissing the unknowns. Many claims fade under measurement. The ones that don’t are more interesting.
  • Produces reusable evidence. Camera poses, 3D reconstructions, and error-barred ranges/speeds can feed physics modeling, radar/IR correlation, and multi-sensor fusion.
  • Aligns with official best practices. AARO, ODNI, and NASA are asking for standardized, quantitative methods; photogrammetry is the bridge between citizens, scientists, and institutions. (AARO)

Acquisition & handling

Metadata & signatures

Photogrammetry

Video & kinematics

Geolocation & timing

Practical debunk-style tests

  • Re-enactment with controlled settings: Same spot, same time of day, stabilized mount, varied focus/ISO/shutter to tease out flare/bokeh/RS effects.
  • Instrumented re-capture: Tripod + star-tracker for night sky; IMU logger for pan rate; dual-camera baseline for instant triangulation.
  • Community replication: Share de-identified data & steps so others can reproduce (or falsify) your result, this is how the strongest Verified cases emerge.

Implications for policy, science, and culture

A UAP archive full of unmeasured anecdotes stalls progress. An archive full of measured, reproducible, geolocated, and time-anchored media transforms the conversation: safety cases get resolved faster; misidentifications are closed with transparent reasoning; and genuine anomalies rise to the top with quantifiable strangeness. This is aligned with the direction from AARO and NASA: more standardized data, fewer ambiguities, and open methods. (AARO)

References 

SEO keywords

UAP photogrammetry, UAP image analysis, UAP metadata forensics, UAP rolling shutter, UAP triangulation, UAP verification, UAP geolocation, UAP satellite flare, UAP camera calibration, AARO UAP imagery, NOAA solar calculator UAP, CelesTrak TLE UAP, COLMAP UAP, Meshroom UAP, OpenCV UAP analysis, ExifTool UAP.

Was this article helpful?
Yes No

Want to contribute?

Are you interested in working with us?
Request to Join